That looks normal.
Do you have access to your YARN ResourceManager WebUI? It should have a 
Scheduler tab on the left, on it you could find out about YARN Application 
Queue. From that you can find out about application pending, memory used and so 
on, perhaps that would help to investigate this further.


    _____________________________
From: Will Du <will...@gmail.com>
Sent: Thursday, December 3, 2015 4:13 PM
Subject: Re: zeppelin job is running all the time
To:  <users@zeppelin.incubator.apache.org>


       All is what’s by default     Properties                     name      
value                args                      master      yarn-client          
      spark.app.name      Zeppelin                spark.cores.max               
       spark.executor.memory      512m                spark.home      
/usr/hdp/current/spark-client/                spark.yarn.jar      
hdfs:///apps/zeppelin/zeppelin-spark-0.5.5-SNAPSHOT.jar                
zeppelin.dep.localrepo      local-repo                zeppelin.pyspark.python   
   python                zeppelin.spark.concurrentSQL      false                
zeppelin.spark.maxResult      1000                zeppelin.spark.useHiveContext 
     true                   
                     On Dec 3, 2015, at 11:51 AM, Felix Cheung <      
felixcheun...@hotmail.com> wrote:          
                           Could you send us your configuration of the Spark 
Interpreter in Zeppelin?                      
                      It can see how both jobs can be long lived in Spark & 
hive/Tez but they should not block one or another.        
        
                
                           _____________________________       
From: Will Du <       will...@gmail.com>       
Sent: Thursday, December 3, 2015 2:15 AM       
Subject: Re: zeppelin job is running all the time       
To: <       users@zeppelin.incubator.apache.org>       
       
       
         When I run spark-shell and hive cml, everything is good.               
  One more thing I find is that when I run the hive CML, there is single hive 
job running (seen from hue job browser) all the time until I exit from hive 
cml.                         No matter how many HQL I submit, it is still 
single job there. I think it is because of using Tez. Is Tez having conflict 
with zeppelin?                         
                                          On Dec 3, 2015, at 2:15 AM, Felix 
Cheung <            felixcheun...@hotmail.com> wrote:                      
                                                              I don't know 
enough about HDP, but there should be a way to check user queue in YARN?        
      
Spark job shouldn't affect Hive job though. Have you tried running spark-shell 
(--master yarn-client) and a Hive job at the same time?              
               
                             From:               will...@gmail.com              
 
Subject: Re: zeppelin job is running all the time               
Date: Tue, 1 Dec 2015 22:56:55 -0500               
To:               users@zeppelin.incubator.apache.org               
               
Do I have to stop zeppelin to make hive job running?                            
   I am thinking to have one notebook have both spark and hive coding running.  
                                            
                                              In addition, I am quite sure the 
first job in the picture is hive job. The 2nd one is zeppelin running spark. I 
wonder how to unblock hive job.                                              
                                                                      On Dec 1, 
2015, at 10:45 PM, Jeff Zhang <                   zjf...@gmail.com> wrote:      
                              
                                                         >>>                    
 the status of hive query becomes from ACCEPT to RUNNING                        
                 The status you see on the hue is the yarn app status which is 
used for spark. It should be running forever until you shutdown it in zeppelin  
                                                                              
                                         On Wed, Dec 2, 2015 at 11:42 AM, Will 
Du                     <will...@gmail.com> wrote:                     
                                                                  The latest 
HDP sandbox 2.3.2                                                               
                                   
                                                                                
                              On Dec 1, 2015, at 10:38 PM, Jeff Zhang <         
                    zjf...@gmail.com> wrote:                                    
                    
                                                                                
       Which version of HDP do you use ?                                        
                                                
                                                             On Wed, Dec 2, 
2015 at 11:23 AM, Will Du                               <will...@gmail.com> 
wrote:                               
                                                                                
                I have assigned an dedicate yarn queue to spark, the status of 
hive query becomes from ACCEPT to RUNNING. However, it sees running forever 
still. Everything is else is in default config of HDP sandbox. Do I need to set 
something else? The browser I saw status is from hue.                           
                                        
                                                                                
                    Thanks,                                                     
                                               wd                               
   
                                                                     
                                                                                
                                                                                
                         On Nov 30, 2015, at 12:40 AM, Rick Moritz <            
                           rah...@gmail.com> wrote:                             
         
                                                                                
                                                                          To 
explain the previous reply: the SparkContext created by Zeppelin is persistent 
and independent of whether it's currently processing a paragraph or not. 
Therefore the Zeppelin job will claim all ressources assigned to it until 
Zeppelin is stopped.                                                            
                                                        
                                                                               
On Mon, Nov 30, 2015 at 4:20 AM, Jeff Zhang <zjf...@gmail.com> wrote:
                                                                                
                                                                                
      I assume this is yarn app Job Browser. How many executors do you specify 
for your zeppelin yarn app ? It seems your zeppelin yarn app consume all the 
resources so that it block other applications.                                  
                                                     
                                                                                
                                                  
                                                                                
                                                                                
                                                                                
               
                                                                                
         On Mon, Nov 30, 2015 at 11:10 AM, Will Du <will...@gmail.com> wrote:
                                                                                
                                                          Hi folks,             
                                                                                
    I am running a simple scala word count from zeppelin in HDP sandbox. The 
job is successful with expected result. But the job of zeppelin is shown in job 
status of hue forever. It seems block the my hive job. Does anyone know why?    
                                                                                
                                                             thanks,            
                                                                                
                                                     wd                         
                                                                                
                                      <PastedGraphic-1.tiff>                    
                                                                                
                                                                                
                                              
                                            
                                                                                
         
                                                                                
                                                                                
           -- 
                                                                                
           Best Regards                                              
                                              
Jeff Zhang                                                                      
                                                                                
                                                       
                                                                                
                                                                                
                     
                                                                                
                                                                                
                              
                              
                                                             
                              --                              
                                                             Best Regards       
                        
                               
Jeff Zhang                                                                      
                                                                                
                
                                                                                
                                                                           
                    
                                         
                    --                    
                                         Best Regards                     
                     
Jeff Zhang                                                                      
                                    
                                                                                
             
                
       
                      
   


  

Reply via email to